Context Awareness in Health Care: a Review

i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 6 ( 2 0 0 7 ) 2–12

journal homepage: www.intl.elsevierhealth.com/journals/ijmi

Review Context awareness in health care: A review

a, b Nathalie Bricon-Souf ∗, Conrad R. Newman a Centre d’Etude et de Recherche en Informatique Medicale´ , Universite´ de Lille2, 59045 Lille, France b Centre for Health Informatics, University of New South Wales, Sydney, Australia a r t i c l e i n f o a b s t r a c t

Article history: Background: Health care systems will integrate new computing paradigms in the coming Received 5 April 2005 years. Context-awareness computing is a research field which often refers to health care as Received in revised form an interesting and rich area of application. 4 January 2006 Aim: Through a survey of the research literature, we intended to derive an objective view Accepted 4 January 2006 of the actual dynamism of context awareness in health care, and to identify strengths and weaknesses in this field. Methods: After discussing definitions of context, we proposed a simple framework to analyse Keywords: and characterize the use of context through three main axes. We then focused on context- Context awareness awareness computing and reported on the main teams working in this area. We described Medical informatics some of the context-awareness projects in health care. A deeper analysis of the hospital- Health care based projects demonstrated the gap between recommendations expressed for modelling context awareness and the actual use in a prototype. Finally, we identified pitfalls encountered in this area of research. Results: A number of opportunities remain for this evolving field of research. We found relatively few groups with such a specific focus. As yet there is no consensus as to the most appropriate models or attributes to include in context awareness. We conclude that a greater understanding of which aspects of context are important in a health care setting is required; the inherent sociotechnical nature of context-aware applications in health care; and the need to draw on a number of disciplines to conduct this research. © 2006 Elsevier Ireland Ltd. All rights reserved.

Contents

1. Introduction ...... 3 1.1. Aim of the paper...... 3 1.2. Methodology used for developing context-awareness analysis ...... 4 2. What is context?...... 4 2.1. Deﬁnitions ...... 4 2.2. Analysis framework ...... 4 3. What is context-awareness computing? ...... 5 4. The “research context” of “context-awareness computing in health care”...... 5

∗ Corresponding author. Tel.: +33 3 20 62 68 29; fax: +33 3 20 62 68 81. E-mail address: [email protected] (N. Bricon-Souf). 1386-5056/$ – see front matter © 2006 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.ijmedinf.2006.01.003 i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 6 ( 2 0 0 7 ) 2–12 3

4.1. Main research teams and their projects ...... 5 4.2. Description of health care context-awareness projects ...... 5 4.2.1. Application ...... 5 4.2.2. Hospital-based prototypes ...... 7 4.2.3. Other prototypes ...... 7 4.2.4. Scenario ...... 7 5. What are the challenges? ...... 7 5.1. Actual characteristics of context-awareness application at the hospital ...... 7 5.2. Identiﬁed pitfalls with the use of context-awareness computing...... 9 5.2.1. The system must remain intelligible ...... 9 5.2.2. People invent new conventions through new tools ...... 9 5.2.3. The purpose of the tool could be deviated ...... 9 6. Discussion ...... 9 7. Conclusion...... 10 References ...... 10

1. Introduction In particular, current technologies allow the introduction of context awareness in every day activities. Health care will evolve as new technologies are adopted. Even Context awareness is a concept that has been described if it is difficult to predict what the future hospital will be, for some time, but technologies (e.g. wireless technologies, aspects such as context awareness will help health care pro- mobile tools, sensors, wearable instruments, intelligent arti- fessionals to shift part of their activities to machines. Rein- facts, handheld computers) are now available to support the vention of health care [1] is complex. In this paper, we aim to development of applications. Such technologies could help outline difficulties and possible solutions in the area of con- health care professionals to manage their tasks while increas- text awareness in health care. ing the quality of patient care. Nevertheless, new technolo- Even with the ongoing increase in hospitals use of com- gies impact the communication between agents. In his paper puterized tools (e.g. powerful hospital information systems, “Interaction design”, Coiera [16] presents a framework for the connected laboratory results) these tools are not sufficient design of interactions between human and computational and new technologies should support a new way of envis- agents working in organizations. He describes the impact of aging the future hospital. The future intelligent hospital will a new interaction class (in this discussion, a class that would be deeply different from the current one. The introduction of include context-awareness applications) within an organiza- future tools leads to challenging research problems [2–9]. First, tion. The introduction of a new interaction class will impact such an evolution requires new technologies and new archi- the level of interaction and communication by agents in terms tectures to implement secure and reliable systems. It requires of costs and benefits to individuals. When designing new inter- the identification and evaluation of what could be done, for action classes, one aim is to ensure optimization of benefit what purpose and how this could be implemented. It could versus cost to individuals. also induce new social or political problems in relation to pri- Health care systems could integrate context-awareness vacy concerns or acceptance of such systems. computing, not only to explore new tools but to propose useful Our focus is on identification, evaluation and implementa- and acceptable systems. tion of new tools or services for the communication and the Intensive care units (ICU’s) contain complex health care sit- cooperation of health care professionals. uations and are a challenging area for such systems. A number Communication between health care professionals repre- of researchers have underlined this context of work as par- sents a large part of their activity [10,11]. This communication, ticularly relevant to the evaluation of complex tools assisting direct or indirect, ranging from laboratory results to com- the cooperation between workers [17–19]. The medium term plex consultation and advice, is important and useful but at perspective of our research is the definition of a set of require- the same time induces many interruptions. This complicates ments for the use of context-awareness tools in the ICU. the cognitive activity of the health care providers. Cooper- Our preliminary work consists of the review of context ation of health care professionals is indispensable to care, awareness in health care, and is presented in this paper. but problems with transmission of information between them still induce breakdowns in communication [12,13]. Errors, 1.1. Aim of the paper which sometimes lead to the death of a patient, have been described in US hospitals [14], and obviously exist in all Through a survey of the existing literature in context- hospitals. awareness computing, we intend to discuss the most relevant Cooperation between health care professionals can be research relating to health care. increasingly mediated through computerized platforms (e.g. Section 2 presents a brief state of art on what is con- hospital-GP intermediation platforms, homecare coordination text, then proposes a framework for the analyses of the platforms, etc.). These coordination tools will integrate new context representation, in order to perform this review mobile tools and propose new communications abilities [15]. of the context-awareness projects in health care. Section 4 i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 6 ( 2 0 0 7 ) 2–12

3 brieﬂy proposes a deﬁnition of context-awareness com- proposed by Dey matches this view. Another view can be sus- puting. Section 4 focuses on context-awareness comput- tained by phenomenological theory—context emerges from ing in health care: it presents some of the more rele- the activity and cannot be described independently. Some vant teams in this area, then describes the main health similar remarks are made by Winograd: “something is context care projects developed in such a topic. Section 5 presents because of the way it is used in interpretation” [23]. Faced with the challenges linked to such project: it analyses existing the complex notion of context, Sato proposes to represent con- research on hospital context awareness for health care and text through “a pattern of behavior or relations among vari- describes how closely such existing prototypes or applica- ables that are outside of the subjects of design manipulation tions match the above recommendations, then it mentions and potentially affect user behavior and system performance” the main pitfalls encountered in context-awareness comput- [24]. ing area. Section 6 gives a summary and a conclusion of this survey. 2.2. Analysis framework

1.2. Methodology used for developing Using those perspectives, we proposed a simple framework to context-awareness analysis analyse the use of context in health care applications, choos- ing three main axes to characterize context. These axes refer, Major electronic research databases (Medline1 through respectively, to the following questions: What context is used PubMed, scientific journals via their own sites or Science for? What are the context items of information? Are the con- Direct) as well as a web search engines (Google predominantly) text features invariant and if no, how is it possible to organize were used to identify research published in the area of context them? awareness (and related fields) and health care. A selection strategy was formulated by the authors focus- Purpose of use of context: • ing on research themes of (i) a proposed general framework According to Dey et al. [21], context is used in three main for context awareness or (ii) “health care AND context aware- cases: (i) presentation of information and services to a user, ness”. Main conferences in this domain were identified and (ii) execution of a service and (iii) tagging of context to infor- reviewed, and related references of the main papers studied. mation for later retrieval. The search was broadened to connected areas of research such Items for context representation: • as ubiquitous or pervasive computing if used in the context of Through the analyses of the health care context-awareness health care. projects, it is possible to describe the items of context used. Synthesising our view of the literature we examined, we identified three main classes to split items of context into: 2. What is context? (i) people, (ii) environment and (iii) activities. Organization of context features: • The concept of context intersects with a diverse range of Recent literature highlights the complexity of the fea- research (e.g. artificial intelligence, ontology, knowledge repre- tures of context. Context representation is not only split- sentation, etc.). We do not intend to perform a complete review ted on the further two axes mentioned above, but should of what is context in this paper, interested reader might refer be organized in more sophisticated ways. Different orga- to Brezillon’s work [20], but we report on definitions, elabo- nizations are then proposed. We identified further refine- rated by people working on context-awareness computing to ment of the organization through consideration of the produce the required framework that could be used for the following: review of the existing researches on this topic. A hierarchical organization, which draws on from gen- ◦ 2.1. Definitions eral to local aspects of context, or from generic to specific aspect [25] “activities and the contexts that surrounds It is not obvious to define context—Dey et al. [21] proposed them do not exists in isolation, but rather in cascade, fol- a definition of context which is the following: “Any informa- lowing the structure of activities from the general to the tion that can be used to characterize the situation of entities specific”. (i.e. whether a person, place or object) that are considered rel- An organization according to the dimension of concept ◦ evant to the interaction between a user and an application, [26]: internal aspect of context (mood of the user, state of including the user and the application themselves. Context the device), or external aspect (temperature, time, etc.). is typically the location identity and state of people, groups An organization according to the focus of the current ◦ and computational and physical objects”. But Dourish [22] activity, leading to consider context with different levels argued that there is an incompatibility between two views of granularity [27]. of context. One comes from positivist theory—context can be An organization according to the current usefulness of ◦ described independently of the actions done; the definition context: manifesting aspect (relevant for current action), or latent (non-relevant for current action) [34].

1 Medline: National Library of Medicine’s premier bibliographic In our analyses of health care project, we mentioned the database covering the ﬁelds of medicine, nursing, dentistry, veteri- existence of such an organization of the context features if it nary medicine, the health care system and the preclinical sciences. exists. i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 6 ( 2 0 0 7 ) 2–12 5

52) papers in January 2005. We must be aware of the weakness 3. What is context-awareness computing? of such simple indicators but it highlights the interest in this area of computing in medical informatics. Context-aware computing can be defined as “an application’s ability to adapt to changing circumstances and respond 4.1. Main research teams and their projects according to the context of use” [45]. This notion is anything but simple, and implementing such This section presents some of the main research teams work- an application will raise challenging issues. We refer to Satya- ing on the topic of context awareness. We summarize different narayanan’s work [27] to present issues linked to context research teams focus in Table 2. For each team, we indicate awareness. We report them, as described in Table 1. trends in their main area of research (e.g. computer sciences Definitions of models and architectures able to support or social sciences); how much they elucidate a definition of such applications are still an area of active research and con- context; their involvement in the definition of an architecture sensus has not been reached. Some problems to be addressed or framework; if they are proposing an application or proto- relate to equipment (e.g. networks, captors, artifacts, mobile type and their involvement in medical applications. communication). Others concern security, especially impor- The reader is referred to Chen and Kotz [28] and Korkea- tant in the medical domain (e.g. identification, authentication, aho [29] for surveys of context-awareness computing research availability, integrity, confidentiality). Our focus is on context- and for a more detailed description of some applications. A awareness research relevant to health care. good overview of the issues in this domain can be found in “Context-Aware Services State-of-the-Art” [30] (WASP Dutch project). 4. The “research context” of Even though most of the general papers on context aware- “context-awareness computing in health care” ness indicate health care as an important and promising field of research, the majority of laboratories we identified are The use of handheld computers has been increasing in many essentially involved in computer sciences research. Health health care applications—Medline, for example, uses “com- care, if it exists, appears as one of their interests for a potential puter handheld”, as a new indexing term of the MeSH the- testing area of the proposed frameworks or tools. saurus. This term was introduced in 2003 and retrieves more Only a few laboratories were identified in the literature as and more papers on this topic: about 250 papers in May 2004 being involved in context awareness dedicated to health care and more than 400 in January 2005. Many researchers focus on research. The Centre for Pervasive Health care, in Denmark, is improving the mobility of health care professionals. completely dedicated to this research and The University of Other new computing paradigms also support research in Irvine examines social aspects of context awareness in health this area. In fact, there are several research themes which care. share similar concepts: context awareness, ubiquitous computing, embedded computing, pervasive computing, sentient 4.2. Description of health care context-awareness computing and others. Even though these terms are well- projects known in the field of computer science, their use in medical applications is quite rare. Nevertheless, this research area The following section will detail some of the context-aware is attractive enough to have spurred workshops or topics as medical applications proposed in the literature. First, we part of international conferences (e.g. “Pervasive Computing describe the context-awareness projects, and then we use the in Health Care”, Pervasive Computing 2002; “Building Bridges: main characteristics of context awareness, as identified earlier Interdisciplinary Context Sensitive Computing and Mobile in this paper, to describe how these projects deal with context. Computing in Medical Context”, Glasgow 2002; “Ubiquitous Computing for Pervasive Health Care Applications”, Ubicomp 4.2.1. Application 2003; and CFP “Pervasive Health Care” in a special issue of 4.2.1.1. Vocera communication system—experimentation in St. the IEEE EMBS journal (end 2003)). A search using Medline on Vincent hospital, Birmingham, USA [48]. The Vocera commu- those keywords is an indicator of the current trends: [Medical nication system is a communicator badge system for mobile Informatics] (mesh term) and “mobile” (respectively, “ubiqui- users. It is a wearable badge with a push-to-call button, a small tous”, then “pervasive”) retrieves 375 (respectively, 135, then text screen and versatile voice-dialing capabilities based on

Table 1 – Issues in context awareness from Satyanarayanan [27] Implementing a context-aware system requires many issues to be addressed. For example: How is context represented internally? How is this information combined with system and application state? Where is context • stored? Does it reside locally, in the network, or both? What are the relevant data structures and algorithms? How frequently does context information have to be consulted? What is the overhead of taking context into account? What • techniques can one use to keep this overhead low? What are the minimal services an environment needs to provide to make context awareness feasible? What are reasonable fallback • positions if an environment does not provide such services? Is historical context useful? What are the relative merits of different location-sensing technologies? Under what circumstances should one be used in preference • to another? Should location information be treated just like any other context information, or should it be handled differently? 6 i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 6 ( 2 0 0 7 ) 2–12 ao anan y; ela; y; Galv y y v a e F Gonzalez Intille; Bates; ic; Redd + ohanson; J g anar ez. v Quigle d anson; x; Mahane cia r o a eople F Sw Saty P isher; w Christensen; Gar Spiteri; F wd. Smaila Nielsen Rodrig Burnett otz Ed ek Burleson. ad; K Kjedsko Lim; USA am; gr ahan Abo Rondoni hell; ce v; arty; dr a wior y; g ino ekop; v om ummerfeld; i r o De Chen; Munoz. Sato; Belloti; R Mitc Selker; K F P Dourish; W Bar Boa. Gonzalez Pier Sie Coulouris Bossen; Sko Hanr Martinez fr sites). b to lot) that e w ant elop (w Medical v (scenario) (one/a their application ++ + +++ + + + +++ +++ de topic) m o r f y + +++ ++ ++ + +++ +++ ++ + + +++ +++ +++ (mainl ototype Pr teams ork the w of me principles a r F + + + +++ +++ ++ + + +++ +++ +++ +++ + +++ Design + +++ description al ener g Ctxt + + + + + + + +++ + + + + + + +++ + + +++ + + +++ ++ to of ed Definition eferr . r , . S ea . S Comp name ar . . . . ., . Social S Info Info Info Info ., ...... oject ocus Comp S S S S S S S S S S S F ...... pr Social , and . Comp Comp Comp Medical Comp Medical Comp Medical Comp Comp Comp Medical S Design Comp Comp Comp ends tr e e v al asi CIT hospital. Gener perv . e ubiquitous orkspaces ojects HealthCar of W ojects: ee pr r Room oolkit oject e in pt futur CfPC p paper T v Computing pr b Hospital; ent the the e acti La f xico this Ctxt v conce o needs oject; hospital” + action-fr for in Me Mobile pr VEPSY acti Inter ARD Intellig ent d Media for distr Home l xtIT cited Some a e en, Inter r y e opean ar eness v oject URA: w ectl ar Se Oxyg Information A A computing QosDREAM “intellig ISIS: Conte computing MobileW Sandfor Pr Merino Center Eur Ensenada, w dir a e , v e h [47] gy c those [46] [23] asi ies e g , ersity of h e v ar erv [37,27] context c g P of [42–44] xico [31–33] hnolo g Milano Resear c ersity b California, Uni in [25] , e v hnolo Me 3 for Cambridg T g c a La C Colle of e ho e of names ersity Resear Mellon T Uni of O v Institute [21,36] Aarhus e d i Pittsbur ends , a r g , i e and gy [28] Alto ersity g T Media [24,38] o-Psyc [40] v Canberr ersity r [39] – , v e [22,34,35,26,8] y go alo v 2 ersity v Neur alia—DST alia—Uni hnolo ences e c l e xique—CICESE, b Irvine T Uni [41] Chica Hano Healthcar [45] Center Sydne Center for ustr ustr a Refer T USA—MIT USA—Uni USA—Geor USA—Carne UK—Uni USA—Stanfor USA—Institute USA—Darmouth Denmark—Centr Denmark—Aalbor Me A A USA—P Italia—Applied i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 6 ( 2 0 0 7 ) 2–12 7 speech recognition. It allows hand free conversation: hand free 4.2.3. Other prototypes answer to calls, and voice message if there is no answer. It The above projects are hospital based. Other specific projects is biometrically secured with speaker verification. It delivers exist, more in the domain of pervasive health care. information directly to the users and avoids the need to go to a distant device (phone, PC). Some deal with medication consumption: Fishkin and Wang [49] propose to assist medication given at 4.2.2. Hospital-based prototypes home: a pad is designed to detect when a bottle of medica- 4.2.2.1. Hospital of the future—Centre for Pervasive Health tion is lifted off and put back, which bottles are moved and care, Denmark [42,43]. A context-aware prototype is proposed, how many pills are removed. Information such as how many including: pills to take, and post-medication suggestions are displayed behind the pad. Floerkemeir and Siegemund [50] propose A context-aware hospital bed with a built-in display which smart blister packs: equipped with sensors, they are able • can be used by patient for entertainment (e.g. for viewing to detect the removal of the pills. The patient will receive television) and by clinicians for accessing medical data. The a reminder on his mobile phone if it is detected that he has bed “knows” who is using it, and what and who is near it. not taking the prescribed medication. A context-aware pill container that is “aware” of the patient, Some deal with distant monitoring: • and reveals itself when near the patient (by lighting the Korhonen et al. [51] propose a social alarm for elderly based proper container with the name of the patient). on wearable sensors (a wrist unit—detection of movement) A context-aware Electronic Patient Record. and intelligent monitoring (IST Vivago System®). • Some are new assistants: The bed “knows” the nurse, the patient and the medicine Helal et al. [52] propose to combine Java smart phones and tray, and displays relevant information according to this con- smart home to provide assistance for elderly patients (mobile text, such as a medicine schema or patient record. patient care giving assistant, general reminder system, aug- mented awareness: notification of events such mail delivery, 4.2.2.2. Intelligent hospital software—University of Cambridge, water leak, etc.). Mihailidis et al. [53] present an assistant UK [41]. Following a study of the needs at the Accident & during hand washing for adults with dementia. Emergency Department of the Royal London Hospital, the authors proposed scenarios of use (remote consultation, track- 4.2.4. Scenario ing of patients and equipment, notification of awareness and 4.2.4.1. Representation of context for a hospital information patient data) and have implemented an experimental pro- system that uses PDAs to deliver patient information to totype. The prototype allows localisation of a team member doctors—Institute of Design, Chicago [38]. The team presents and the ability to initiate an audio–video conference from the a simple example in a medical setting to show how context nearest point. Clinicians are localised, are notified of the call could be described by combining contextual information from and can acknowledge the call through their active badge. This the hierarchy of environment (hospital room), person (patient) prototype is presented as a demonstrator of the middleware and activity (doctor–patient status discussion). platform QoS DREAM, for reconfigurable multimedia stream- ing and event-based programming. 5. What are the challenges? 4.2.2.3. Context-aware mobile communication—CICESE, Mex- ico [46]. The idea of this project is to empower mobile devices Context-awareness computing is still challenging. In this sec- to recognize the context in which hospital workers perform tion, we focus on three major aspects of such challenges: their tasks. In particular the authors propose an extension of instant messaging to add context awareness as part of the The first one is that we did not find any recommendations message such as circumstances that must be satisfied before about the functional needs of the context. The second one the system delivers the message. Contextual elements used concerns the gap which still exists between fundamental include location, delivery timing, role reliance, artifact (partic- researches on context representation and actual context- ularly the device) location and state. A prototype is proposed, awareness prototypes. We present these points through a which is able to provide contextual messaging, for example, “a deeper analysis of the hospital-based projects mentioned message for room 226 to any physician, delivery time for the above. message today after 2 p.m.”. The last point deals with the difficulties in building efficient computerized systems for a mediation of human perspec- 4.2.2.4. MobileWARD—Mobile Electronic Patient Record, Aal- tives. We develop that point through a presentation of some borg University, Denmark [45]. MobileWARD is a prototype pitfalls mentioned in literature and encountered in context- designed to support morning procedure tasks in a hospital awareness computing. ward, and is able to display patients lists and patient information. The device presents information and functionality 5.1. Actual characteristics of context-awareness according to the location of the nurse and the time of the application at the hospital day. This project simulates the context events linked to the location of the staff member. Patients are chosen through a Table 3 summarizes the characteristics of the hospital-based patient-list or an activation of a barcode at the bed-side. projects described above. The second column describes the 8 i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 6 ( 2 0 0 7 ) 2–12 e r list e xtual a xtual v instant affect w of h ates acti gnition system conte xt-a conte clinical alerts, pills alert oject sub-system ener of e eco whic d, r r g on pr ough a with h conte the with w , speaker thr ecor of the r ing EPR audio–video y unication ence g r Description , xt-a speec e e use v i influences l e Comm Location Conte Mobile Handheld patient messa d badg confer Combination information of the with (dialing), identification container bed information a y xt onl anized but conte g s Is or e No Y No scenario No No No and es oom timing r artefacts time location onment patient vir featur ulated En Location Staff artifact Hospital Medication, location Sim location Location artifact—location es ole ole identity ofessional , ID/r ID/r ID r featur p , patient ID attending of eliance r eople ole P ysician ofessional ofessional ofessional r atient atient ecipient Pr P P Role Pr ID/r status r P ph e ar es artifact) status people featur ysical state ph what vity of (on Acti s o e N Idea doing Doctor–patient Y Artifact discussion A, A e vice een PD PD de scr badg ing video ojects , g e e artefacts v v pr artifact A , acti PD mobility mobility mobile messa Settings speaker sensiti eness o h r printer a w Hospital micr touc Hospital PC Hospital PC, instant Hospital Hospital b i i Use Pi Pi Pi P P context-a a S S S S P P P ype T M A P M M M hospital the , v of e hospital eam & in yses mobile 2003) anson, partment , unicating (Kjeldsko e am) e r QosDr a D (Sw dr Anal video ince Healthcar w y hell) – e ARD e (V ork (Bar Comm Accident v M 3 unication enc xt-influence xt-a w go) a g Mitc w asi xico) v) oject ame able ocer erv ollo V System F application, FmWk Emer (UK, Conte fr Chica Sko MobileW T P Pr Conte comm (Me i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 6 ( 2 0 0 7 ) 2–12 9

type of work done (proposed model (M), prototype (P), appli- sonalization, passive context awareness and active context cation (A) or scenario (S)). The third column describes the use awareness. They study how participants evaluate these three of context according to the first axe of categorization pre- levels of interaction and conclude: “users are willing to accept sented in Section 2.2 (presentation of information (Pi), exe- a large degree of autonomy from applications as long as the cution of services (ES), storage of contextual information (SI)). application’s usefulness is greater than the cost of limited con- Columns 5–7 are used to describe the different features of control”. text manipulated by the system, in the three main classes described for the second axe of Section 2.2, i.e. activity, people 5.2.2. People invent new conventions through new tools and environment. The eighth column refers to the third axe of Brown and Randell [54] analysed how people, “dwelling” with Section 2.2 and describes if the context is organized through technology, use a phone ring or medical alerts. First, they a categorization. The last column describes briefly the objec- show that suggesting correct behavior using context (even tives of use of context awareness. for such simple tools) is incredibly complex. It is perhaps Most of these projects are prototypes and real applications more useful to use context defensively, to avoid incorrect are still difficult to find. This is obviously due to the difficulty behavior. Secondly, they show that people use technology to of managing a distributed and complex system for context communicate context to users, for example, ringing another awareness. If we look more precisely at the context, we can person twice indicates that the call is urgent; a very recent remark that: call means that the caller is still next to his phone, and should reply when called back. Their recommendation is to We can find explanations on how the situations are man- provide simple structures so that users could adopt and use • aged, but it is more difficult to know “why” such settings technology. have been chosen and how it meets functional needs. Con- sidering medical work in hospital, papers give some general 5.2.3. The purpose of the tool could be deviated recommendations (for example, Bardram mentions that From a human perspective, designing tools which are able context awareness is particular useful for user-interface to achieve good communication is not an obvious task. Fog- navigation, communication systems are also underlined by arty et al. [37] have designed a context-aware system which different teams). is able to indicate the availability of an individual, taking The features used are often very simple—time, location and into account their location, computer and calendar infor- • identity of health care staff member and patient are the mation, in the hope it could help people avoid disruptions most common attributes described. through interruptions. They report that in testing, partici- No organization of context is used even if recommended in pants do not experience a reduction in interruptions, rather • the more fundamental studies of context awareness. they use the context provided to identify the presence of Activity knowledge is used infrequently. the person they want to speak to, not to check his/her • availability. One of the most advanced projects is the hospital of the future proposed by Bardram [43], in which activity is treated by using the artifact (bed, pill containers) to inform about the 6. Discussion current actions of medical staff. Cooperation of actors in health care requires a large amount of 5.2. Identified pitfalls with the use of diverse information types. Of the potential set of candidates context-awareness computing for context feature, there is room for more complete explo- ration (e.g. elements from the patient record, global activities, Acceptance of a system at the individual, social, political or etc.). The authors agree that there are specific features unique economic level will strongly influence its future use. Different to the “rhythm” of health care work (particularly in a hospi- studies illustrate that even if context awareness is an attrac- tal) that while considered by some authors [35] have yet to be tive proposition, one must be careful when using such appli- explored to a sufficient level of granularity. Many researchers cations. We identified in our literature review problems which discuss the importance of location as an important attribute could occur and difficulties linked with human perspectives: in context-aware systems, but it is the author’s view that this may not have the same priority as a requirement of such sys- 5.2.1. The system must remain intelligible tems in a health care setting. In contrast activity is felt by Belloti and Edward [39] argue that humans have complex moti- the authors to be a crucial component of communication but vations which lead to unpredictable behaviors. The presence even if cited, it is rarely used with the importance afforded of context-aware systems, inferring human intent and medi- location. ating between people, without intelligibility and accountabil- This review illustrates the reality that there remain many ity, could be problematic. The prototype proposed by Kjeldskov questions in the field of context awareness in health care, with and Skov [45], which only takes into account location for the few answers as yet to the questions/issues described in the display of Electronic Patient Record on nurses’ PDA, strength- research of Satyanarayanan [27]. ens this point “some nurses became confused or even annoyed It is apparent to the authors that developing context-aware by the automatic adaptation of information on the screen to applications is not a problem of access to technologies (sen- their physical location”. Barkhuus and Dey [36] describe three sors, networks, etc.) as these exist. One difficulty is, as yet, levels of interactivity with context-awareness computing: per- the research community has not reached a consensus as to 10 i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 6 ( 2 0 0 7 ) 2–12

the best way to model context and architectures to support its use. Summary points Also no preferred direction for the use of context awareness “What was already known before the study was done” has emerged, with at present a range of potential uses (instant messaging, video conferencing, etc.) evident in the literature. Papers mentioned the complexity of context. • It is likely that the evaluation of such systems in health care Papers mentioned the usefulness of context- • settings will guide the focus of future research, as some can- awareness computing. didate applications are found to have negative impact such as Papers mentioned “health care” as interesting settings • increased communication load on health care workers or an for context-awareness computing. increase in errors. But no requirements about: • Despite the often cited claim that health care is an ideal which interesting health care settings can use ◦ candidate for context-awareness applications there seems to context-awareness computing; be a paucity of actual systems development and implementa- which interesting context features can be used to ◦ tion in this area. Even if one considers prototypes, a large gap implement health care applications. exists between the requirements expressed in the literature we reviewed and the prototypes developed to date. This is not “What this study has added to our knowledge” surprising however given that this is a new area of research in an early stage of its evolution. An overview of the existing systems for health care. • An overview of the features of context actually used. • An analysis of the gap between actual context- • 7. Conclusion awareness prototypes in health care and the modelling of contextual applications proposed in the literature. The fact that no preferred contextual needs emerge In this paper we have tried to provide an overview of the new • area of context awareness in health care. While we have not yet. covered all aspects of research in this area, or all projects, we have attempted to provide an overview of this evolving area, identify which areas of research are more dynamic and to highlight the gaps that exist. In order to classify the research r e f e r e n c e s reviewed, we developed a matrix to illustrate our key conclusions. Bear in mind that the conclusions we drew were based on the author’s opinion and are not meant to be a quantita- [1] E. Coiera, Four rules for the reinvention of health care, BMJ tive assessment of the various research groups’ strengths in 328 (2004) 1197–1199. different areas. [2] K. Lyytinen, Y. Yoo, Issues and challenges in ubiquitous It seems that this area remains a fertile area for research. computing, Commun. ACM 45 (12) (2002) 63–65. Despite the opinion by many research groups that health care [3] G. Davis, Anytime/anyplace computing and the future of would be an ideal focus area for context-aware applications knowledge work computing, Commun. ACM 45 (12) (2002) 67–73. there are relatively few medical informatics research groups [4] A. Fano, A. Gershman, The future of business services in the actively involved in such research. In order to develop this age of ubiquitous computing, Commun. ACM 45 (12) (2002) research area a greater understanding of which aspects of con- 83–87. text are most important in the health care domain is required. [5] J. Grundin, Groups dynamics and ubiquitous computing, Assumptions about the relative importance of an attribute Comput. Commun. ACM 45 (12) (2002) 74–77. such as location are yet to be validated by the evaluation of [6] D. Siewiorek, New frontiers of application design computing, Commun. ACM 45 (12) (2002) 79–82. systems beyond limited clinical trials. It is when such applica- [7] L. Jessup, D. Robey, The relevance of social issues in tions are integrated into the complex network of interactions ubiquitous computing environments computing, Commun. within a hospital that the value (or otherwise) of a design will ACM 45 (12) (2002) 88–91. become evident. Given that error is such a concern in the [8] G. Banavar, A. Bernstein, Software infrastructure and design health care setting [14], we also need to be mindful of the challenges for ubiquitous computing applications, Comput. potential for new applications areas to introduce a new set Commun. ACM 45 (12) (2002) 92–96. of errors. [9] J. Bohn, F. Gartner¨ , H. Vogt, Dependability issues of pervasive computing in a healthcare environment, in: Proceedings of As a ﬁnal point, Coiera [1] argues that since health systems the First International Conference on Security in Pervasive are sociotechnical systems, where outcomes emerge from the Computing (SPC), Lecture Notes in Computer Science, interaction of people and technologies, we cannot design orga- Springer-Verlag, 2003. nizational or technical systems independently of each other. [10] E. Coiera, When conversation is better than computation, J. Lorenzi [57] articulates a number of non-technical issues that Am. Med. Inf. Assoc. 7 (2000) 277–286. can act as barriers to a new systems success. [11] R. Spencer, P. Logan, Role-based communication patterns The development of such systems will need to draw on within an emergency department setting, in: Proceedings HIC 2002, Melbourne, August 2002. expertise from a number of disciplines including (but not lim- [12] R. Beuscart, V. Carre, F. Westrelin, S. Dalmas, et al., Interim ited to) computer science, social sciences, design and evalu- Report on Co-operative Work in CCE (Methodology), ation. Our hope is that this review will stimulate discussion AIM-TANIT (A2036) Project, COOP Workpackage, Deliverable and aid research in this area. D2, Public Report, 1992. i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 6 ( 2 0 0 7 ) 2–12 11

[13] M. Grosjean, M. Lacoste (Eds.), Communication et [35] M. Reddy, P. Dourish, A finger on the pulse: temporal intelligence collective: le travail a` l’hopital,ˆ P.U.F., Le Travail rhythms and information seeking in medical work, in: Humain, Paris, 1999. Proceedings of the ACM Conference on Computer Supported [14] L.T. Kohn, J.M. Corrigan, M.S. Donaldson (Eds.), To Err is Cooperative Work CSCW2002, 2002. Human: Building a Safer Health System, 2000. [36] L. Barkhuus, A. Dey, Is context-aware computing taking [15] N. Bricon-Souf, M.C. Beuscart-Zephir, R. Beuscart, control away from user? Three levels of interactivity Communication of Information in the Homecare Context in examined, in: Proceedings of Ubicomp 2003, Springer, 2003, Processing of the MIE Conference, St. Malo, France, May 4–7, pp. 150–156. 2003. [37] J. Fogarty, J. Lai, J. Christensen, Presence versus availability: [16] E. Coiera, Interaction Design Proceedings of the IMIA WG the design and evaluation of a context-aware HIS Working Conference, Heidelberg, April 2002. communication client, Int. J. Hum. Comput. Stud. 61 (3) [17] N. Bricon-Souf, J.M. Renard, R. Beuscart, Dynamic workflow (2004) 299–317. model for complex activity in intensive care unit, Int. J. Med. [38] E. Swanson, A. Galvao, K. Sato, A framework for Inf. 53 (1999) 143–150. understanding contexts in interactive systems [18] J.M. Renard, N. Bricon-Souf, J.M. Geib, R. Beuscart, Simulation development, in: 7th World Multi-Conference on Systemics, of dynamics tasks routing to improve cooperation in Cybernetics and Informatics, Orlando, FL, July 2003. intensive care units, Stud. Health Technol. Inf. (1999) [39] V. Belloti, K. Edwards, Intelligibility and accountability: 31–36. human considerations in context aware systems, Hum. [19] M. Reddy, W. Pratt, P. Dourish, M. Shabot, Sociotechnical Comput. Interact. 16 (2–4) (2001) 193–212. requirements analysis for clinical system methods, Inf. Med. [40] B. Kummerfeld, A. Quigley, C. Johnson, R. Hexel Merino, 4 (2003) 437–444. Towards an intelligent environment architecture for [20] P. Brezillon, Context in problem solving: a survey. Knowl. multi-granularity context description, in: Proceedings of Eng. Rev. 14 (1999) 1–34. Workshop on User Modelling for Ubiquitous Computing, [21] A. Dey, G. Abowd, D. Salber, A conceptual framework and Pittsburg, June 2003. toolkit for supporting the rapid prototyping of context-aware [41] S. Mitchell, M. Spiteri, J. Bates, G. Coulouris, Context aware applications in special issue on context-aware computing, multimedia computing in the intelligent hospital, in: Hum. Comput. Interact. J. 16 (2–4) (2001) 97–166. Proceedings of the Ninth ACM SIGOPS European Workshop, [22] P. Dourish, Seeking a foundation for context aware Denmark, September 2000. computing, Hum. Comput. Interact. 16 (2) (2001) 229–241. [42] J. Bardram, Applications of context-aware computing in [23] T. Winograd, Architectures for context, Hum. Comput. hospital work—examples and design principles, in: Interact. 16 (2) (2001) 401–419. Proceedings of SAC, Cyprus, March 14–17, 2004. [24] K. Sato, Context sensitive interactive systems design: a [43] J. Bardram, Hospitals of the future—ubiquitous computing framework for representations of contexts, in: Proceedings support for medical work, in: Hospitals Workshop Ubihealth of the 10th International Conference on Human–Computer 2003, 2003, Interaction, vol. 3, Hum.-Centered Comput. 3 (2003) http://www.healthcare.pervasive.dk/ubicomp2003/papers/. 1323–1327. [44] J. Bardram, C. Bossen, A. Lykke-Olesen, K. Haslov Madsen, R. [25] P. Prekop, M. Burnett, Activities, context and ubiquitous Nielsen, Virtual video prototyping of pervasive healthcare computing, Comput. Commun. 26 (2003) 1168–1176. systems, in: Symposium on Designing Interactive Systems, [26] P. Dourish, What we talk about when we talk about context, 2002, pp. 167–177. Pers. Ubiquitous Comput. 8 (1) (2004) 19–30. [45] J. Kjeldskov, M. Skov, Supporting work activities in [27] M. Satyanarayanan, Pervasive computing: vision and healthcare by mobile electronic patient records, in: challenges, in: IEEE Personal Communication, 2001. Proceedings of the 6th Asia–Pacific Conference on [28] G. Chen, D. Kotz, A Survey of Context-Aware Mobile Human–Computer Interaction, APCHI 2004, Rotorva, New Computing Research, Dartmouth Computer Science Zealand, 2004. Technical Report TR2000-381, Hanover, 2000. [46] M. Munoz, M. Rodriguez, J. Favela, A. Martinez-Garcia, V. [29] M. Korkea-aho. Context-Aware Applications Survey, 2000, Gonzalez, Context-aware mobile communication in http://www.hut.fi/ mkorkeaa/doc/context-aware.html. hospitals, IEEE Comput. 36 (9) (2003) 38–46. ∼ [30] S. Pokraev, P.D. Costa, J.G. Pereira Filho, M. Zuidweg, J.W. [47] G. Riva, Ambient intelligence in health care, Cyberpsychol. Koolwaaij, M. van Setten, Context-Aware Services Behav. 6 (3) (2003) 295–300. State-of-the-Art, Report, WASP Project, Nederland, [48] V. Stanford, Beam me up, Dr. McCoy, IEEE Pervasive Comput. November 2003, pp. 1–86, Mag. 2 (3) (2003) 13–18. https://doc.telin.nl/dscgi/ds.py/Get/File-27859. [49] K. Fishkin, M. Wang, A Flexible, Low-Overhead Ubiquitous [31] S.S. Intille, L. Bao, E. Munguia Tapia, J. Rondoni, Acquiring in System for Medication Monitoring, Intel Research Seattle situ training data for context aware ubiquitous computing Technical Memo IRS-TR-03-011, October 25, 2003, applications, in: Proceedings of CHI 2004 Connect: http://seattleweb.intel- Conference on Human Factors in Computing Systems, ACM research.net/people/fishkin/pubs files/medpad tr.pdf. Press, April 2004. [50] C. Floerkemeier, F. Siegemund, Improving the effectiveness [32] T. Selker, W. Burleson, Context-aware design and interaction of medical treatment with pervasive computing in computer systems, IBM Syst. J. 39 (3–4) (2000) 880–891. technologies, in: Ubihealth, 2003, [33] S.S. Intille, J. Rondoni, C. Kukla, I. Iacono, L. Bao, A http://www.healthcare.pervasive.dk/ubicomp2003/papers/. context-aware experience sampling tool, in: Proceedings of [51] I. Korhonen, P. Paavilainen, A. Sar¨ ela,¨ Application of the Conference on Human Factors and Computing Systems, ubiquitous computing technologies for support of 2003. independent living of the elderly in real life settings, in: [34] D. Fisher, P. Dourish, Social and temporal structures in Ubihealth, 2003, everyday collaboration, in: Proceedings of the SIGCHI http://www.healthcare.pervasive.dk/ubicomp2003/ Conference on Human Factors in Computing papers/. Systems, Table of Contents, Vienna, Austria, 2004, pp. [52] S. Helal, C. Giraldo, Y. Kaddoura, C. LEE, H. El Zabadani, W. 551–558. Mann, Smart phone based cognitive assistant, in: Ubihealth, 12 i n t e r n a t i o n a l j o u r n a l o f m e d i c a l i n f o r m a t i c s 7 6 ( 2 0 0 7 ) 2–12

2003, http://www.healthcare.pervasive.dk/ubicomp2003/ [54] B. Brown, R. Randell, Building a context sensitive telephone: papers/. some hopes and pitfalls for context sensitive computing, in: [53] A. Mihailidis, B. Carmichael, J. Boger, G. Fernie, An intelligent Proceedings of the Joint Colloquia: Building Bridges: environment to support aging-in-place, safety, and Interdisciplinary Context Sensitive Computing and Mobile independence of older adults with dementia, in: Ubihealth, Computing in Medical Context, Glasgow, September 9–10, 2003, http://www.healthcare.pervasive.dk/ 2002. ubicomp2003/papers/. [57] N. Lorenzi, Beyond the gadgets, BMJ 328 (2004) 1146–1147.